Snoop Dogg sensual seduction

pho's picture

Hi,

Any clue on how to do this?

https://vimeo.com/23497797

harrisonpault's picture
Re: Snoop Dogg sensual seduction

Not certain what M.O. did, but it looks very much as though: Used a Kinect for 1) rgb feed (top left) 2) depth map feed (top right) 3) detects the edge of the depth map 4) puts it on a 12-cell Queue 5. Renders all 12 items from the queue in an iterator, using a growing scale and rainbow color based on queue position (main output)

Without a Kinect, the top right might use some other CV technique to find the figure in the rgb feed.

Here is a little prototype of the composition structure, using a particle system as the source image...I don't have a Kinect.

PreviewAttachmentSize
pth expanding image queue.qtz10.64 KB

dust's picture
Re: Snoop Dogg sensual seduction

@MOostrik use's his open cl depth select with the kinect in combination with open cl edge detection etc.. very useful macros.

@MOostrik Depth Select - http://kineme.net/forum/Applications/Compositions/KinectDepthSelectopenCL

see attached edge detect macro @MOostrik .

the trace / trail effect in the video is more than likely done with open cl knowing @MOostrik. here are two projects where trail effects are deployed in open cl. http://kineme.net/composition/MOostrik/AugmentedRealityinOpenCL http://kineme.net/composition/MOostrik/OpenCLBoids

i modified the open cl queue structure patch to take float4's which you could use to stack images much like the above post but in open cl. http://kineme.net/composition/dust/parallelpainting

the other routes would be the classic memo / hybrid trace effect which would maybe be easiest to get up and running. im thinking just take the depth select and run it into this accumulation and core image kenrel to produce the snoop effect. http://www.hybridvisuals.nl/downloads/m8-trace-quartz-composer-effect/

PreviewAttachmentSize
Simple_Edge_Detection.qtz1.83 KB
Depth_Select_Dust.qtz23.92 KB